Improving the fairness and robustness of deep learning classification models through bias mitigation and post hoc calibration

Abstract Purpose: The goal of this study was to enhance the fairness and clinical applicability of deep learning models for classifying Alzheimer’s disease from structural MRI scans. To achieve this, we addressed two key challenges: age-related bias and unreliable probability estimates. Specifically, we investigated in-processing debiasing strategies to achieve age-invariant predictions while preserving diagnostic accuracy. We also systematically evaluated post hoc calibration methods to improve the reliability, interpretability, and clinical usefulness of predicted probabilities across diverse patient populations. Methods: Structural MRI data from three independent cohorts (ADNI, AIBL, and DELCODE) were preprocessed using a standardized pipeline. Age-related bias was addressed using two in-processing strategies, adversarial training (CF-Net) and GLM-based feature residualization (PMDN), implemented in a shared 3D convolutional neural network (CNN) framework. Additionally, post hoc calibration was performed using four regression methods: logistic, probit, isotonic, and Bayesian binning. We evaluated model performance, fairness, and calibration using classification as well as calibration-specific metrics. Results: PMDN achieved the best overall performance, with higher balanced accuracy (0.71) and improved calibration (Brier $$\approx $$ ≈ 0.22), showing substantially reduced variability compared to baseline and CF-Net. PMDN effectively removed age-related bias, achieving near-zero age prediction correlation; CF-Net, however, showed only partial and unstable debiasing. Post hoc calibration further improved probability reliability, with nonparametric methods (isotonic regression and Bayesian binning) yielding the lowest calibration errors (MCE $$\approx $$ ≈ 0.21−0.24). Conclusion: This study shows that, although deep learning models achieve similar classification performance, they differ significantly in terms of calibration, stability, and fairness. PMDN provides more consistent and effective age bias mitigation than the CF-Net approach, particularly in clinically relevant subgroups. Post hoc calibration improves probability reliability further, with nonparametric methods achieving the best calibration.

Authors

Institutions

Publication Details

Journal
International Journal of Computer Assisted Radiology and Surgery
Published
2026-10-17
DOI
https://doi.org/10.1007/s11548-026-03805-x
Primary Topic
Dementia and Cognitive Impairment Research
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Improving the fairness and robustness of deep learning classification models through bias mitigation and post hoc calibration

Majid Ramedani, Ayda Rostamzadeh, Lukas Preis, Tobias Strauß et al.
International Journal of Computer Assisted Radiology and Surgery
Dementia and Cognitive Impairment Research
article

Improving the fairness and robustness of deep learning classification models through bias mitigation and post hoc calibration

Majid Ramedani, Ayda Rostamzadeh, Lukas Preis, Tobias Strauß, Klaus Fließbach, Martin Dyrba, Enise Irem Incesoy, Julian Hellmann‐Regen, Anja Schneider, Wenzel Glanz, Sebastian Sodenkamp, Falk Lüsebrink, Peter Dechent, Sophia Stöcklein, Eike Jakob Spruth, Maria Gemenetzi, Marie Theres Kronmüller, Ole H. Martensen, Josef Priller, Annika Spottke, Björn H. Schott, Jaya C. Terli, Jens Wiltfang, Emrah Düzel, Stefan Hetzer, Oliver Peters, Boris-Stephan Rauchmann, Katharina Buerger, Christoph Laske, Stefan J. Teipel, Daniel Janowitz, Robert Perneczky, Frank Jessen
article en

Abstract

Abstract Purpose: The goal of this study was to enhance the fairness and clinical applicability of deep learning models for classifying Alzheimer’s disease from structural MRI scans. To achieve this, we addressed two key challenges: age-related bias and unreliable probability estimates. Specifically, we investigated in-processing debiasing strategies to achieve age-invariant predictions while preserving diagnostic accuracy. We also systematically evaluated post hoc calibration methods to improve the reliability, interpretability, and clinical usefulness of predicted probabilities across diverse patient populations. Methods: Structural MRI data from three independent cohorts (ADNI, AIBL, and DELCODE) were preprocessed using a standardized pipeline. Age-related bias was addressed using two in-processing strategies, adversarial training (CF-Net) and GLM-based feature residualization (PMDN), implemented in a shared 3D convolutional neural network (CNN) framework. Additionally, post hoc calibration was performed using four regression methods: logistic, probit, isotonic, and Bayesian binning. We evaluated model performance, fairness, and calibration using classification as well as calibration-specific metrics. Results: PMDN achieved the best overall performance, with higher balanced accuracy (0.71) and improved calibration (Brier $$\approx $$ ≈ 0.22), showing substantially reduced variability compared to baseline and CF-Net. PMDN effectively removed age-related bias, achieving near-zero age prediction correlation; CF-Net, however, showed only partial and unstable debiasing. Post hoc calibration further improved probability reliability, with nonparametric methods (isotonic regression and Bayesian binning) yielding the lowest calibration errors (MCE $$\approx $$ ≈ 0.21−0.24). Conclusion: This study shows that, although deep learning models achieve similar classification performance, they differ significantly in terms of calibration, stability, and fairness. PMDN provides more consistent and effective age bias mitigation than the CF-Net approach, particularly in clinically relevant subgroups. Post hoc calibration improves probability reliability further, with nonparametric methods achieving the best calibration.

International Journal of Computer Assisted Radiology and Surgery
University of Bonn (DE), University of Cologne (DE), University Hospital Bonn (DE), German Center for Neurodegenerative Diseases (DE), Munich Cluster for Systems Neurology (DE), Universitätsmedizin Göttingen (DE), UK Dementia Research Institute (GB), Universitätsmedizin Rostock (DE), Cologne Excellence Cluster on Cellular Stress Responses in Aging Associated Diseases (DE), University Hospital Magdeburg (DE), Hertie Institute for Clinical Brain Research (DE), University of Rostock (DE), Imperial College London (GB), University of Aveiro (PT), Technical University of Munich (DE), University of Göttingen (DE), Charité - Universitätsmedizin Berlin (DE), University of Tübingen (DE), Ludwig-Maximilians-Universität München (DE), University of Sheffield (GB), Otto-von-Guericke-Universität Magdeburg (DE), University of Edinburgh (GB)
Alzheimer's Disease Neuroimaging Initiative, Deutsche Forschungsgemeinschaft, Deutsches Zentrum für Neurodegenerative Erkrankungen, National Institutes of Health
Good health and well-being
Openalex Percentile: Top 12%
Dementia and Cognitive Impairment Research
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.